566 Mr. M. Solomon on the 



foundation-pillar, it was found impossible to get good readings 

 when the traffic was at all frequent. The w4iole set of obser- 

 vations was therefore made between twelve and six on the 

 morning of June 23rJ, 1898. A sixty-ohm simple reflecting 

 galvanometer was used with an astatic combination of four 

 small needles affixed to a wire to the upper end of which the 

 mirror was fastened, the whole being suspended by a single 

 silk fibre. This system swung between the coils of the galva- 

 nometer, and as it had no mica vane the air damping was not 

 great. As the coils were of fine wire, and the parts of the 

 galvanometer nearly all of wood, the eddy-currents would be 

 very small when the coils were on open circuit, and the 

 damping in that case would be almost entirely due to the air 

 resistance and the viscosity of the suspension. When, how- 

 ever, the galvanometer circuit was closed by short-circuiting 

 the terminals, eddy-currents would be set up, and would 

 introduce a considerable factor in the damping; so that if the 

 alteration in the ratio tfk is due to the factor introduced by 

 the eddy-currents, it would be observable with the galvano- 

 meter circuit closed, but not with it open. 



The determinations of the period were made by observing 

 with a stop watch the time taken in executing several com- 

 plete swings — from five or six w T hen the period was long, to 

 twenty when it was short. The time was read as the spot of 

 light passed the zero of the galvanometer scale. Unfortu- 

 nately in the experiments with a period of 16 seconds the 

 true zero was about 25 divisions out, in such a direction as to 

 cause the observed period to be too small. Had the ampli- 

 tudes of the first and last swings used in determining the 

 period been observed, it would be possible to make an accurate 

 correction : as this was not done, it is only possible to make 

 an approximate one, which will not, however, be far wrong, 

 and which has been made in the results given in the table. 



In observing the decrement when the period was long, the 

 amplitudes of successive swings on both sides of the zero were 

 observed. If these be a, b, c, d, e, . . . <fcc, we have 



Decrement = 7 =-=... = = — — -j = . . . ac. 



b c b + c c + d 



By thus using: -, instead of j- to calculate the decrement 



6 b + c b 



any zero error is eliminated. A specimen of the actual 



readings obtained in an observation of this type is appended. 



